Fix make release for 4-STABLE

[Scott Long]

John Baldwin wrote:
> On Monday 19 January 2004 05:58 am, Makoto Matsushita wrote:
> 
>>list> What about splitting the floppies? Splitfs support is already on
>>list> the 4.9-RELEASE disks.
>>
>>That would be a viable option, but the great downside of using splitfs
>>is that all of splitted chunks are required for all users.
>>
>>In the current configuration, drivers.flp is an option; if you don't
>>need to use modules in drivers.flp, you only need two floppies.  If
>>the kernel and/or mfsroot image are splitted into, say, three
>>floppies, we have to have three floppies.
>>
>>Yes, floppies are so cheap to buy, but sometimes not handy you know.
>>I believe that some of my friends said to me that "hey, I'm so tired
>>to write THREE floppy images, it's so slow!"
>>
>>It would be technically easy to introduce splitfs for kernel/mfsroot
>>(it is already there), but it also introduce an important design change.
>>We should have enough time to consider that we decide the 3rd floppy
>>image treats a mandatory one.
> 
> 
> I have splitfs done in a p4 tree.  Currently for i386 on current it takes up 3 
> floppies.  One more floppy image is not all that bad, and it removes all the 
> kernel module complication.  This is using a GENERIC kernel, so it greatly 
> simplifies release building and requires no more tweaking of BOOTMFS to make 
> stuff "just fit".  I'm currently doing some final testing before committing 
> it to HEAD as I've occasionally had problems with the split images being 
> loaded off of floppy.  I added a md5 command to the loader today and it 
> verified that the mfsroot and kernel were read ok, however.  I have this bad 
> feeling that there is some kind of memory corruption bug in the loader and 
> that the problem goes away if you use a loader that has forth in it.
> 

This is excellent work.  Thanks a lot.

Scott